Big Data – Creating New Underwriting Opportunities VNAB Algemeen... · LATLONG WS STREET WS FULL...
Transcript of Big Data – Creating New Underwriting Opportunities VNAB Algemeen... · LATLONG WS STREET WS FULL...
0
Big Data – Creating New Underwriting Opportunities VNAB 17th March 2015
1
Agenda
• Introduction
• Increased competition
• Importance of accurate data
• A growth opportunity through increased insight
2
The latest in a long line of buzzwords
3
Beyond the buzz
• Identify new opportunities, deliver new service
• Improve internal efficiency • Strategic overview, cross sell to existing
clients
• Consider the accuracy of data
Data types and sources
Meteorological data
Mixture of internal and external data sources
Property
Motor
Health
Census data
Real estate data - Kadaster
Credit history
Location data – mobiles / GPS
Social media – Facebook, Twitter, Google+
Sales records
Claims history
Probabilistic / historic modelling
HR records
Financials
• Partnerships and alliances • Access to external data • Skills and capabilities
Telematics
Wrist bands
4
Increased competition
5
Growth in reinsurer capital continues to out-pace demand, rising by 6% to a new high of USD575bn at September 30, 2014
$575bn of reinsurance capital chasing $320bn of demand Interest in directing capital towards Insurance arena increasing Additional parties such as Google+ entering insurance space
Competition increasing
17 22 19 22 24 28 39 50 62
368 388
321 378
447 428 466
490 513
6% -17% 18%
18%
-3% 11%
7%
6%
385 410
340
400
470 455
505 540
575
0
100
200
300
400
500
600
FY 2006 FY 2007 FY 2008 FY 2009 FY 2010 FY 2011 FY 2012 FY 2013 9M 2014
USD
(bill
ions
)
Traditional CapitalAlternative CapitalGlobal Reinsurer Capital
USD 23bn
USD6bn USD4bn
USD 29bn
Catastrophe Bonds
Sidecars Industry Loss Warranties
Collateralised Reinsurance
Alternative Capital
6
Opportunity increasing
0
50
100
150
200
250
300
350
400
450
500
1980 1985 1990 1995 2000 2005 2010
USD
bill
ions
(201
4 pr
ices
)
Insured Loss (2014 USD) Economic Loss (2014 USD)Source: Impact Forecasting (all figures inflation-adjusted)
258 catastrophe events in 2014
USD 132billion economic loss USD 39billion insured loss
Increased risk understanding delivers new opportunities, such as Netherlands Flood
7
Importance of accurate data
8
Property opportunity Increasing risk understanding – which data? Physical and financial information related to a risk
General Information
Primary Risk Characteristics
Secondary Modifiers
Financial Information
Sums Insured
Geographic Location
Coverage Type
Occupancy Construction Year Built
Number of Storeys
Roof Shape and Construction
Building Shape
Deductibles Limits
Co-insurance
9
Property opportunity Current quality – UK vs Netherlands
Sums Insured
Detailed Geographic
Location
Coverage Type
Detailed Occupancy
Construction Year Built
Number of Storeys
Roof Construction Building Shape
Deductibles Limits
Co-insurance
Sums Insured
Detailed Geographic
Location
Coverage Type
Detailed Occupancy
Construction Year Built
Number of Storeys
Roof Shape and Construction
Building Shape
Deductibles Limits
Co-insurance
What is driving this quality mis-match? Influence of internet buying? Impact of Agency underwriting data sources?
UK Netherlands
10
Catastrophe modelling – Risk insight AIR (Applied Insurance Research) (Verisk Analytics business unit) founded in 1987; over 80 countries covered
EQECAT (acquired by CoreLogic in 2013) founded in 1994; over 90 countries covered
RMS (Risk Management Solutions) (DMGT group company) founded in 1988; over 80 countries covered
Impact Forecasting Aon Benfield subsidiary 50+ available models Originally, to cover gaps in peril coverage available from vendor models and emerging perils Now, the main stake is on transparency and bespoke solutions
11
Property opportunity Importance of accurate location data
AAL
250
1000
LATLONG FLSTREET FL
FULL POSTCODE FL4 DIGIT FL
CRESTA FL
Influence of Geographic Resolution on River Flood Loss Ratio
AAL
250
1000
AAL
250
1000
LATLONG WSSTREET WS
FULL POSTCODE WS4 DIGIT WS
CRESTA WS
Influence of Geographic Resolution on Windstorm Loss Ratio
AAL
250
1000
Coordinate Street address
6 digit Postcode
4 digit postcode City CRESTA
12
Client A Modelled losses for Germany river flood doubled in 2013 without any obvious change in exposure Investigation revealed – the same industrial risk was recorded with postcode resolution in 2012 data
and was refined to exact address in 2013 …
Real case – how important it is to get this right!
Actual risk is located here
2012
2013
13
Inaccurate construction data could impact underwriting price, capital requirement and loss potential by 50%+ in above example
Property opportunity Importance of accurate construction data
-100%
-50%
0%
50%
100%
150%
NL_Wood frame NL_Masonry NL_Reinforced masonry NL_Reinforced concreteImpa
ct o
n Lo
ss E
stim
ate
Territory and Construction
Impact of Construction (NL Commercial) - 1 in 200 year loss
AIREQERMS v11RMS v9
14
Height of property especially important for flood but can significantly impact windstorm loss potential as well
Property opportunity Importance of accurate height data
-100.00%
-80.00%
-60.00%
-40.00%
-20.00%
0.00%
20.00%
40.00%
NL_1 Storey NL_2 Storey NL_5 Storey NL_10 Storey NL_20 Storey
Impa
ct o
n Lo
ss E
stim
ate
Territory and Storeys
Impact of Storeys (NL Residential) - 1 in 200 year loss
AIREQERMS v11RMS v9
15
Our objective is to support you in developing complete, accurate and appropriate exposure data
Benchmarking We will conduct benchmarking studies to advise on accuracy and quality of data compared to peers
Data enhancement We can further enhance the key data characteristics of your data, such as building height and size, using our access to third party building resolution dataset for the Netherlands. This data can help to further refine your windstorm and flood PML’s
Property - Increasing risk understanding through partnership
16
Property – Benefits of increased insight
Combining enhanced data with catastrophe modelling will lead to more accurate original pricing, PMLs, reinsurance requirements and enhanced capital efficiency
Additionally we can identify locations that are under or over weight in your portfolio
Generate modelled burning costs (annual average loss) by peril by location to identify locations that are lowest or highest exposed to loss potential
Target current areas of low exposure and low loss potential for growth
17
A growth opportunity through increased insight
18
Property – Netherlands flood – a new opportunity for growth 1st market available probabilistic model Model characteristics Both probabilistic and scenario model
Overtopping and/or failure of primary defenses of 57 dyke rings
Covers dyke failure mechanisms such as overflow, wave overtopping, piping, inward macro instability, damage to cover and erosion of levee core, dune erosion and failure of hydraulic structures (overflow/overtopping, failure to close, seepage, structural failure)
Not considering precipitation or sewer system overload
Covers building, content, BI, motor hull
Modeling aspects NL heavily affected by man-made modifications Cannot be modeled in a ‘common’ way Small amount of historic loss data External cooperation with local experts vital (HKV and Deltares) and local
sources of information used
19
Netherlands flood – Event set One million years of events simulated
Probabilistic and historic event modelling
Each event can be visualised and interrogated
Model can be bespoked to fit specific portfolio characteristics
Events producing floods are stored in the event set
Event ID Occurence DR1 DR2 DR3 DR4 DR5 DR6 DR7 … DR11 DR12330 5137_S1 2.08E-06 1 1 0 0 1 219 0 … 0 71
In event 330, there is flood in 5 dyke rings (1,2,5,6,12). In dyke ring 6 scenario 219 applies (water source is Wadden Sea), in dyke ring 12 scenario 71 applies (breach location near Holwerd). Right map shows affected postal codes.
20
Netherlands flood – Visualising and adjusting for Uncertainty The Impact Forecasting model has 3 main views of potential flood extent by line of business
(including motor casco – Default, Optimistic and Pessimistic
Additional Uncertainty options can also be explored within the Elements platform to identify impact of data accuracy:
10 25 50 75 100 150 200 250 500 1000
Commercial Realistic Commercial Optimistic Commercial Pessimistic
21
Big data - Conclusion
The influx of alternative capital and data rich companies such as Google+ will lead to increased competition and further pricing pressures Thinner margins means the ability to accurately price risks and identify profitable growth more important than ever
The insurance industry is sitting on a wealth of data and unparalleled market expertise The new capital is already harnessing data analytics but lacks underwriting and market expertise
Could the use of social media generated data be biased and / or less than accurate? Do specific types of people use and post data online?
Partnerships will be key to data enhancements, modelling and more accurate risk assessment Ability to draw down on outputs for use in targeting growth opportunities, improved pricing and more accurate capital and
reinsurance needs
22
© 2013 Aon Benfield Netherlands cv Reinsurance Brokers Alle rechten voorbehouden. Niets uit deze rapportage mag worden verveelvoudigd, opgeslagen in een geautomatiseerd gegevensbestand, of openbaar gemaakt, in enige vorm of op enige wijze, hetzij elektronisch, mechanisch, door fotokopieën, opnamen, of op enige andere manier, zonder voorafgaande schriftelijke toestemming van Aon Benfield Netherlands cv Reinsurance Brokers.
Aon Benfield Netherlands Paalbergweg 2-4 1105AG Amsterdam Nederland